Although many studies have addressed the statistical properties of P coda, there have been few attempts to analyze coda in a deterministic framework. In this analysis, 1050 teleseismic P-wave seismograms from 32 Yucca Flat explosions are inverted for the locations of any isotropic point scatterers in the source region that contribute coherent arrivals to the coda. A rectangular grid of potential scattering locations is constructed around the source array and the semblance (a measure of waveform ensemble coherence) for each point is computed by slant-stacking the full set of seismograms at various moveout velocities. The potential “resolution”, or sensitivity, of the method for this data set is tested by constructing a corresponding suite of synthetic seismograms with direct arrivals plus arrivals from various point scatterers and calculating the semblance at each point on the grid. The particular source-station geometry produces better sensitivity near the array. Sensitivity is enhanced parallel to the trend of the elongated source array and diminished in the orthogonal direction. Stacking the data for a velocity of 2.5 km/sec yields relatively high semblance values in the central portion of the test site as well as about 10 km to the ENE, and moderately high semblance values about 10 km to the WNW. The latter two regions are associated with topographic highs. Simulations using synthetics with random arrivals in the coda indicate that incomplete suppression of random noise may explain the high semblance values in the central region, but probably not in the outlying regions. However, simulations of a circular distribution of random scatterers within 10 km of the test site produce high semblance about 10 km to the ENE, similar to that seen for the data, arising from variable sensitivity. Thus, two possible explanations for the high semblance values are scattering of higher mode surface waves to P waves off topographical irregularities, or random scattering due to heterogeneity in the immediate area of the test site.